Random Graphs and The Parity Quantifier Phokion G. Kolaitis Swastik Kopparty UC Santa Cruz MIT & & IBM Research-Almaden Institute for Advanced Study

2 What is finite model theory? It is the study of logics on classes of finite structures. Logics: First-order logic FO and various extensions of FO: Fragments of second-order logic SO. Logics with fixed-point operators. Finite-variable infinitary logics. Logics with generalized quantifiers.

3 Main Themes in Finite Model Theory Classical model theory in the finite: Do the classical results of model theory hold in the finite? Expressive power of logics in the finite: What can and what cannot be expressed in various logics on classes of finite structures. Descriptive complexity: computational complexity vs. uniform definability (logic-based characterizations of complexity classes). Logic and asymptotic probabilities on finite structures 0-1 laws and convergence laws.

4 Logic and Asymptotic Probabilities Notation:  Q: Property (Boolean query) on the class F of all finite structures  F n : Class of finite structures with n in their universe   n : Probability measure on F n, n ¸ 1   n (Q) = Probability of Q on F n with respect to  n, n ¸ 1. Definition: Asymptotic probability of property Q  (Q) = lim  n ! 1 (Q) (provided the limit exists) Examples: For the uniform measure  on finite graphs G:   (G contains a triangle) = 1.   (G is connected) = 1.   (G is 3-colorable) = 0.   (G is Hamiltonian) = 1.

5 0-1 Laws and Convergence Laws Question: Is there a connection between the definability of a property Q in some logic L and its asymptotic probability? Definition: Let L be a logic  The 0-1 law holds for L w.r.t. to a measure  n, n ¸ 1, if  (  ) = 0 or  (  ) = 1, for every L-sentence .  The convergence law holds for L w.r.t. to a measure  n, n ¸ 1, if  (  ) exists, for every L-sentence .

6 0-1 Law for First-Order Logic Theorem: Glebskii et al. – 1969, Fagin – 1972 The 0-1 law holds for FO w.r.t. to the uniform measure on the class of all finite graphs. Proof Techniques:  Glebskii et al. Quantifier Elimination + Counting  Fagin Transfer Theorem: There is a unique countable graph R such that for every FO-sentence , we have that  (  ) = 1 if and only if R ² . Note:  R is Rado’s graph: Unique countable, homogeneous, universal graph; it is characterized by a set of first-order extension axioms.  Each extension axiom has asymptotic probability equal to 1.

7 FO Truth vs. FO Almost Sure Truth First-Order Truth Testing if a FO-sentence is true on all finite graphs is an undecidable problem (Trakhtenbrot ) Almost Sure First-Order Truth Testing if a FO-sentence is almost surely true on all finite graphs is a decidable problem; in fact, it is PSPACE-complete (Grandjean ). Everywhere true (valid) Somewhere true & Somewhere false Everywhere false (contradiction) Almost surely true Almost surely false

8 Three Directions of Research on 0-1 Laws 0-1 laws for extensions of FO w.r.t. the uniform measure. 0-1 laws for FO on restricted classes of finite structures  Partial Orders, Triangle-Free Graphs, … 0-1 laws on graphs under variable probability measures.  G(n,p) with p ≠ ½ (e.g., p(n) = n -(1/e) )

9 0-1 Laws for Extensions of First-Order Logic Many generalizations of the original 0-1 law, including: Blass, Gurevich, Kozen – Law for Least Fixed-Point Logic (LFP)  Captures Connectivity, Acyclicity, 2-Colorability, … K … and Vardi – Law for Finite-Variable Infinitary Logics L k 1 !, k ¸ 2  Proper extension of LFP K… and Vardi – 1987, Laws for fragments of Existential Second-Order Logic  Capture 3-Colorability, 3-Satisfiability, …

10 Logics with Generalized Quantifiers Dawar and Grädel – 1995  0-1 Law for FO[Rig], i.e., FO augmented with the rigidity quantifier.  Sufficient condition for the 0-1 Law to hold for FO[Q], where Q is a collection of generalized quantifiers. Kaila – 2001, 2003  Sufficient condition for the 0-1 Law to hold for L k 1 ! [Q], k ¸ 2, where is a collection of simple numerical quantifiers.  Convergence Law for L k 1 ! [Q], k ¸ 2, where is a collection of certain special quantifiers on very sparse random finite structures.

11 A Barrier to 0-1 Laws All generalizations of the original 0-1 law are obstructed by THE PARITY PROBLEM

12 The Parity Problem Consider the property Parity = “there is an odd number of vertices” For n odd,  n (Parity) = 1 For n even,  n (Parity) = 0 Hence,  (Parity) does not exist. Thus, if a logic L can express Parity, then even the convergence law fails for L.

13 First-Order Logic + The Parity Quantifier Goal of this work: Turn the parity barrier into a feature. Investigate the asymptotic probabilities of properties of finite graphs expressible in FO[ © ], that is, in first-order logic augmented with the parity quantifier ©.

14 FO[ © ]: FO + The Parity Quantifier © Syntax of FO[ © ]: If  (v) is a formula, then so is © v  (v). Semantics of © v  (v):  “the number of v’s for which  (v) is true is odd” Examples of FO[ © ]-sentences on finite graphs:  © v 9 w E(v, w) The number of vertices of positive degree is odd.  : 9 v © w E(v, w) There is no vertex of odd degree, i.e., The graph is Eulerian.

15 Vectorized FO[ © ] Syntax: If  (v 1,…,v t ) is a formula, then so is © (v 1,…,v t )  (v 1,…,v t ) Semantics of © (v 1,…,v t )  (v 1,…,v t ):  “there is an odd number of tuples (v 1,…,v t ) for which  (v 1,…,v t ) is true” Fact: © (v 1,…,v t )  (v 1,…,v t ) iff © v 1 © v 2  © v t  (v 1,…,v t ).  Thus, FO[ © ] is powerful enough to express its vectorized version.

16 The Uniform Measure on Finite Graphs Let G n be the collection of all finite graphs with n vertices  The uniform measure on G n:  If G 2 G n, then pr n (G) = 1/ 2 n(n-1)/2  If Q is a property of graphs, then pr n (Q) = fraction of graphs in G n that satisfy Q. An equivalent formulation The G(n, 1/2)-model:  Random graph with n vertices  Each edge appears with probability ½ and independently of all other edges

17 FO[ © ] and Asymptotic Probabilities Question: Let  be a FO[ © ]-sentence. What can we say about the asymptotic behavior of the sequence pr n ( , n ¸ 1 ?

18 Asymptotic Probabilities of FO[ © ]-Sentences Fact: The 0-1 Law fails for FO[ © ] Reason 1 (a blatant reason): Let  be the FO[ © ]-sentence © v (v = v) Then  pr 2n (   = 0  pr 2n+1 (  = 1. Hence,  lim n ! 1 pr n (  does not exist.

19 Reason 2 (a more subtle reason): Let  be the FO-sentence © v 1, v 2, …, v 6 Fact (intuitive, but needs proof): lim n ! 1 pr n (  = 1/2 Asymptotic Probabilities of FO[ © ]-Sentences v1v1 v5v5 v6v6 v4v4 v3v3 v2v2

20 Modular Convergence Law for FO[ © ] Main Theorem: For every FO[ © ]-sentece , there exist two effectively computable rational numbers a 0, a 1 such that  lim n ! 1 pr 2n (  a 0  lim n ! 1 pr 2n+1 (  a 1. Moreover, a 0, a 1 are of the form s/2 t, where s and t are positive integers.

21 In Contrast Hella, K …, Luosto LFP[ © ] is almost-everywhere-equivalent to PTIME. Hence, the modular convergence law fails for LFP[ © ]. Kaufmann and Shelah For every rational number r with 0 < r < 1, there is a sentence  of monadic second-order logic such that lim n ! 1 pr n (  r.

22 Modular Convergence Law Main Theorem: For every FO[ © ]-sentece , there exist two effectively computable rational numbers a 0, a 1 of the form s/2 t such that  lim n ! 1 pr 2n (  a 0  lim n ! 1 pr 2n+1 (  a 1. Proof Ingredients:  Elimination of quantifiers.  Counting results obtained via algebraic methods used in the study of pseudorandomness in computational complexity.  Functions that are uncorrelated with low-degree multivariate polynomials over finite fields.

23 Counting Results – Warm-up Notation: Let H be a fixed connected graph. #H(G) = the number of “copies” of H as a subgraph of G = |Inj.Hom(H,G)| / |Aut(H)|. Basic Question: What is pr(#H(G) is odd), for a random graph G? Lemma: If H is a fixed connected graph, then for all large n, pr n (#H(G) is odd) = 1/2 + 1/2 n. Proof uses results of Babai, Nisan, Szegedy – 1989.

24 Counting Results – Subgraph Frequencies Definition: Let m be a positive integer and let H 1,…,H t be an enumeration of all distinct connected graphs that have at most m vertices.  The m-subgraph frequency vector of a graph G is the vector freq(m,G) = (#H 1 (G) mod2, …, #H t (G) mod2) Theorem A: For every m, the distribution of freq(m,G) in G(n,1/2) is 1/2 n -close to the uniform distribution over {0,1} t, except for #K 1 = n mod2, where K 1 is.

25 Quantifier Elimination Theorem B: The asymptotic probabilities of FO[ © ]-sentences are “determined” by subgraph frequency vectors. More precisely: For every FO[ © ]-sentence , there are a positive integer m and a function g: {0,1} t ! {0,1} such that for all large n, pr n (G ² , g(freq(m,G))=1) = 1-1/2 n.

26 Quantifier Elimination Theorem B: The asymptotic probabilities of FO[ © ]-sentences are “determined” by subgraph frequency vectors. Proof: By quantifier elimination. However, one needs to prove a stronger result about formulas with free variables (“induction loading device”). Roughly speaking, the stronger result asserts that: The asymptotic probability of every FO[ © ]-formula  (w 1, …, w k ) is determined by: Subgraph induced by w 1, …, w k. Subgraph frequency vectors of graphs anchored at w 1, …, w k.

27 Notation: type G (w 1, …, w k ) = Subgraph of G induced by w 1,…,w k Types(k) = Set of all k-types freq(m,G,w 1, …, w k ) = Subgraph frequencies of graphs anchored at w 1, …, w k Quantifier Elimination w1w1 w2w2

28 Quantifier Elimination Theorem B’: For every FO[ © ]-formula  w 1, …, w k ), there are a positive integer m and a function h: Types(k) x {0,1} t ! {0,1} such that for all large n, pr n ( 8 w (G ²  (w), h(type G (w ), freq(m,G w))=1)) = 1-1/2 n. Note:  k = 0 is Theorem B.   w 1, …, w k ) quantifier-free is trivial: determined by type.

29 Modular Convergence Law for FO[ © ] Theorem A: For every m, the distribution of freq(m,G) in G(n,1/2) is 1/2 n -close to the uniform distribution over {0,1} t, except for #K 1 = n mod2, where K 1 is. Theorem B: For every FO[ © ]-sentence , there are a positive integer m and a function g: {0,1} t ! {0,1} such that for all large n, pr n (G ² , g(freq(m,G))=1) = 1-1/2 n. Main Theorem: For every FO[ © ]-sentence , there exist effectively computable rational numbers a 0, a 1 of the form s/2 t such that  lim n ! 1 pr 2n (  a 0  lim n ! 1 pr 2n+1 (  a 1.

30 Modular Convergence Law for FO[Mod q ] Theorem: Let q be a prime number. For every FO[Mod q ]-sentece , there exist effectively computable rational numbers a 0, a 1, …,a q-1 of the form s/q t such that for every i with 0 · i · q-1, lim n ´ i mod q, n ! 1 pr n (  a i.

31 Open Problems What is the complexity of computing the limiting probabilities of FO[ © ]-sentences?  PSPACE-hard problem;  In Time(2 2 … ). Is there a modular convergence law for FO[ Mod 6 ]? More broadly,  Understand FO[ Mod 6 ] on random graphs.  May help understanding AC 0 [Mod 6 ] better. Modular Convergence Laws for FO[ © ] on G(n, n -a )?